Bad Models Mistook Housing Bust for Dot-Com Bubble

May 2 (Bloomberg) -- In a speech last month about the financial crisis, Federal Reserve Chairman Ben S. Bernanke trenchantly noted that the initial losses from the dot-com bust were about the same size as those from the housing meltdown -- yet the two episodes had very different economic consequences.

What Bernanke didn’t say was that the reason the Fed, along with every other official forecaster, underestimated the depth of the latest downturn so badly is that its models effectively treated the housing collapse as if it were merely dot-com bust 2.0. And only modest progress has been made toward avoiding that same mistake in the future.

Bernanke said in his speech on April 13, at a conference sponsored by the Russell Sage Foundation and the Century Foundation, that “any theory of the crisis that ties its magnitude to the size of the housing bust must also explain why the fall of dot-com stock prices just a few years earlier, which destroyed as much or more paper wealth -- more than $8 trillion -- resulted in a relatively short and mild recession and no major financial instability.”

He then pointed to the concentration of losses in the financial industry from the housing bust, as opposed to the broad dispersion of fallout from the bursting of the dot-com bubble, as the “principal explanation of why the busts in dot-com stock prices and in the housing and mortgage markets had such markedly different effects.”

The Fed chairman is right: The housing crisis was much more damaging because the initial impact was concentrated in a highly leveraged financial sector and then substantially amplified as those losses cascaded.

Imperfect Models

The problem is that the macroeconometric models used by the Fed -- like those used by the Congressional Budget Office, the White House and others -- had at best a very rudimentary financial sector built into them. As a result, they took into account the macroeconomic impact from the housing bust -- but for the most part didn’t reflect the concentrated loss of wealth and degree of leverage in the financial industry.

In other words, the official models effectively ignored the very distinction that Bernanke highlighted as being crucial to distinguishing the housing collapse from the tech bust. And so the models completely missed the recession’s severity.

In late 2007, for example, the midpoint of the range that the Fed projected for real gross-domestic-product growth in 2008 was more than 2 percent. Instead, real GDP declined by more than 3 percent that year. In early 2008, the Fed projected the economy would expand 2.4 percent in 2009. Instead, real GDP fell 0.5 percent. Those are big forecast errors: The 70 percent confidence interval for one-year-ahead projections is plus or minus 1.2 percentage points. The 2007 projection for 2008 was off by more than five percentage points.

The Fed was far from alone in being overly optimistic; every formal macroeconometric model that I’m aware of made the same mistake. In early 2008, for example, forecasts for 2009 from both the Congressional Budget Office (which I ran at the time) and the Bush administration were roughly in line with, and therefore just as wrong, as the Fed’s.

A paper delivered at the same conference where Bernanke spoke last month suggests one way to address the problem. The economists Simon Gilchrist of Boston University and Egon Zakrajsek of the Fed staff augmented a traditional macroeconomic model with a measure of stress in the financial industry: credit spreads on bonds issued by financial institutions. Those spreads barely moved following the tech bust, but they widened substantially after the housing collapse.

Accounting for Stress

This model, the authors conclude, “can account for the broad movements in consumption, investment, hours worked and output observed during this period.” And since 2007, some progress has been made in incorporating metrics of financial stress into macroeconometric models. But the progress remains inadequate.

So here’s a rough rule of thumb: Whenever a reasonable financial-stress index, such as that produced by Gilchrist and Zakrajsek, is particularly elevated, be very skeptical of economic forecasts from models that pay scant attention to the financial industry. They will be making the housing-meltdown-is-just-like-the-tech-one mistake all over again.

Perhaps fiscal and monetary policy were unaffected by the substantial forecast errors made at the beginning of this crisis. In other words, perhaps if the model predictions had been more dire, the response to the crisis would have been no different. But I doubt it.

(Peter Orszag is vice chairman of global banking at Citigroup Inc. and a former director of the Office of Management and Budget in the Obama administration. The opinions expressed are his own.)

Read more opinion online from Bloomberg View.

Today’s highlights: the View editors on the upside of Greek dysfunction and the U.S. Senate’s attempt to reform the Postal Service; Margaret Carlson on the Romney campaign’s ad strategy; William Pesek on China’s slowing growth; Clive Crook on the dogma of inequality.

To contact the writer of this article: Peter Orszag at

To contact the editor responsible for this article: Mary Duenwald at

    Before it's here, it's on the Bloomberg Terminal.